* High System Load
@ 2004-07-28 23:00 Huntress Gary B NPRI
2004-07-28 23:23 ` Kanoa Withington
0 siblings, 1 reply; 4+ messages in thread
From: Huntress Gary B NPRI @ 2004-07-28 23:00 UTC (permalink / raw)
To: 'linux-raid@vger.kernel.org'
Hi,
I'm creating my first software RAID array. I'm on a 900 MHz system with 786MB of ram, running Fedora RC1. I'm creating a RAID 1 array consisting only of two drives, split on two IDE channels.
I used mdadm --create and everything is running fine so far, but my system load is well over 10. Is this normal? I suspect that this is just overhead while creating the array the first time. Can I assume that when the array is actually put into use that the system loads will not be this high? (I hope not!)
Thanks,
Gary H.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: High System Load
2004-07-28 23:00 High System Load Huntress Gary B NPRI
@ 2004-07-28 23:23 ` Kanoa Withington
2004-07-28 23:54 ` Mike Hardy
0 siblings, 1 reply; 4+ messages in thread
From: Kanoa Withington @ 2004-07-28 23:23 UTC (permalink / raw)
To: Huntress Gary B NPRI; +Cc: 'linux-raid@vger.kernel.org'
There is a lot of activity after you first create a mirror but the CPU
load should not be that high. Maybe you don't have DMA turned on?
Check using "hdparm".
-Kanoa
On Wed, 28 Jul 2004, Huntress Gary B NPRI wrote:
> Hi,
>
> I'm creating my first software RAID array. I'm on a 900 MHz system with 786MB of ram, running Fedora RC1. I'm creating a RAID 1 array consisting only of two drives, split on two IDE channels.
>
> I used mdadm --create and everything is running fine so far, but my system load is well over 10. Is this normal? I suspect that this is just overhead while creating the array the first time. Can I assume that when the array is actually put into use that the system loads will not be this high? (I hope not!)
>
> Thanks,
>
> Gary H.
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: High System Load
2004-07-28 23:23 ` Kanoa Withington
@ 2004-07-28 23:54 ` Mike Hardy
2004-07-29 0:17 ` Jim Paris
0 siblings, 1 reply; 4+ messages in thread
From: Mike Hardy @ 2004-07-28 23:54 UTC (permalink / raw)
To: Kanoa Withington
Cc: Huntress Gary B NPRI, 'linux-raid@vger.kernel.org'
Even without DMA there should only ever be one kernel thread issuing
instructions, resulting in a load of 1, which is all I've ever seen.
DMA vs. PIO modes will just result in a lower resync throughput, which
you can watch with something like (you can even try using hdparm to
change modes while the resync is happening to see the effect if you're
brave):
#!/bin/tcsh
while 1
cat /proc/mdstat
sleep 1
clear
end
There's got to be something else happening on the machine...
-Mike
Kanoa Withington wrote:
> There is a lot of activity after you first create a mirror but the CPU
> load should not be that high. Maybe you don't have DMA turned on?
> Check using "hdparm".
>
> -Kanoa
>
> On Wed, 28 Jul 2004, Huntress Gary B NPRI wrote:
>
>
>>Hi,
>>
>>I'm creating my first software RAID array. I'm on a 900 MHz system with 786MB of ram, running Fedora RC1. I'm creating a RAID 1 array consisting only of two drives, split on two IDE channels.
>>
>>I used mdadm --create and everything is running fine so far, but my system load is well over 10. Is this normal? I suspect that this is just overhead while creating the array the first time. Can I assume that when the array is actually put into use that the system loads will not be this high? (I hope not!)
>>
>>Thanks,
>>
>>Gary H.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: High System Load
2004-07-28 23:54 ` Mike Hardy
@ 2004-07-29 0:17 ` Jim Paris
0 siblings, 0 replies; 4+ messages in thread
From: Jim Paris @ 2004-07-29 0:17 UTC (permalink / raw)
To: Mike Hardy
Cc: Kanoa Withington, Huntress Gary B NPRI,
'linux-raid@vger.kernel.org'
> #!/bin/tcsh
> while 1
> cat /proc/mdstat
> sleep 1
> clear
> end
.. or just 'watch cat /proc/mdstat' :)
> There's got to be something else happening on the machine...
I find that if the kernel variable dev.raid.speed_limit_max is too
high (the default, 200000, is usually too high), then the rebuild
could taking all available I/O time. This can cause every other
process trying to do I/O on the machine to block, which may be the
reason your load average is so high.
Try doing 'sysctl -w dev.raid.speed_limit_max=10000' and see if things
improve.
-jim
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2004-07-29 0:17 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-07-28 23:00 High System Load Huntress Gary B NPRI
2004-07-28 23:23 ` Kanoa Withington
2004-07-28 23:54 ` Mike Hardy
2004-07-29 0:17 ` Jim Paris
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).